http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Intelligent and Real-Time Smart Card Fraud Detection for Optimized Industrial Decision Process
Simeon Okechukwu Ajakwe,Cosmas Ifeanyi Nwakanma,Dong-Seong Kim,Jae Min Lee 한국통신학회 2021 한국통신학회 학술대회논문집 Vol.2021 No.6
Smart Card fraud cases are on the rise due to increased online and real-time untact transactions triggered by several cashless policies as well as the global COVID-19 pan-demic. This paper evaluates various machine learning algorithms for timely and intelligent detection and predictions of smart card frauds. The test the accuracy of the model, dataset of various frauds on smart card was used. Timely, accurate, and intelligent interception of dynamic fraud patterns is crucial to curtail loss rising from security breach. Evaluation results shows that Deep Neural Network (99.91%), Convolutional Neural Network (99.92%), Artificial Neural Network (99.93%) performed better than XGBoost (97.20%), Random Forest (96.12%), Support Vector machine (96.36%), Logistics Regression (96.34%), K-Nearest Neighbour (95.07%), and Naive Bayes (94.87%) both in accuracy and precision. The deployment this deep learning algorithm will guarantee faster smart card fraud detection, improved users’ trust in smart card technology, help in fraud curtailment, facilitate timely strategic counter response measure, as well as trigger further research on improving smart card security, amongst others.
Lightweight CNN Model for Detection of Unauthorized UAV in Military Reconnaissance Operations
Simeon Okechukwu Ajakwe,Rubina Akter,Dong-Seong Kim,Jae Min Lee 한국통신학회 2021 한국통신학회 학술대회논문집 Vol.2021 No.11
In any warfare, success is a function of innovative tactics and strategic deployment of limited resource. This paper develops a lightweight deep neural network that can track and disarm illegal invasion of a territory by drones using radio frequency technology. The dataset consist of RF signals generated from 17 drones at different instances and sources. The result shows that the proposed model achieved a high a prediction accuracy and sensitivity of 95% than existing CNN (80%) and DNN (75%) with low computational complexity.
Simeon Okechukwu Ajakwe,Vivian Ukamaka Ihekoronye,Dong-Seong Kim,Jae Min Lee 한국통신학회 2022 한국통신학회 학술대회논문집 Vol.2022 No.2
The security of key and critical infrastructures is crucial for uninterrupted industrial process flow needed in strategic management as these facilities are major targets of invaders. In this paper, a lightweight vision-based framework for detecting and identifying multi-drones and attached objects was proposed using a deep convolution neural network. The model is validated using 5460 drone samples from six drones and 855 attached objects samples of seven classes under different environments, scenarios (blur, scales, low illumination), and heights. The result reveals that YOLOv5s showed high multi-drone detection of 99.5% and 83.4% payload identification with less time and computational complexity better other YOLOv5 variants which makes it effective for industrial facility aerial surveillance and safety against illegal drone intrusion.
Lightweight CNN Model For Real Time Recognition of Miniaturize Fleet of UAVs
Vivian Ukamaka Ihekoronye,Simeon Okechukwu Ajakwe,Dong-Seong Kim,Jae Min Lee 한국통신학회 2022 한국통신학회 학술대회논문집 Vol.2022 No.2
The accelerating deployment of UAVs in different sectors of life is a plausible contribution in today’s technologically driven world. However, there is also a corresponding increase in UAVs misapplications resulting to economic losses. Therefore, the need to design an efficient anti-drone system in the optimal recognition of drones so as to prevent them from gaining access to restricted areas is a constant strive of researchers. This paper proposes a vision based anti-drone model for optimal detection of miniaturized fleet of drones even in clustered environments. Validation of the proposed model was achieved using 1000 samples of three different drones and birds on several weather scenarios of cloudy, clumsy and sunny conditions. Performance of the proposed model was compared with three other state-of-the-art models based on Precision, Recall and F1-score. Result of simulation shows the superiority of the proposed model in achieving F1-score of 97%, Recall and Precision values of 100% and 94.3% respectively.
UAV Assisted Real-Time Detection and Recognition of Citrus Disease in Smart Farm using Deep Learning
Vivian Ukamaka Ihekoronye,Simeon Okechukwu Ajakwe,Dong-Seong Kim,Jae Min Lee 한국통신학회 2021 한국통신학회 학술대회논문집 Vol.2021 No.11
Timely detection of plant disease is pivotal in a smart farm. This paper presents a deep learning approach for realtime visual detection and classification of citrus disease using a drone. You Look Only Once version 5 (YOLOv5) model was used and tested against the dataset classified as blackspot, canker, greening, and healthy leaves. The result shows that Yolov5 model has superior detection and classification of disease in real-time over Yolov4.